Stretching DNA Using the Electric Field in a Synthetic Nanopore

The voltage threshold coincides with the stretching transition that occurs in double-stranded ... stretches from 107% to nearly 170% of the length of ...
0 downloads 0 Views 487KB Size
Download PDF
NANO LETTERS

Stretching DNA Using the Electric Field in a Synthetic Nanopore

2005 Vol. 5, No. 10 1883-1888

Jiunn B. Heng,† Aleksei Aksimentiev,† Chuen Ho,† Patrick Marks,† Yelena V. Grinkova,‡ Steve Sligar,‡ Klaus Schulten,† and Gregory Timp*,† Beckman Institute and Department of Biochemistry, UniVersity of Illinois, Urbana, Illinois 61801 Received June 8, 2005; Revised Manuscript Received July 29, 2005

ABSTRACT The mechanical properties of DNA over segments comparable to the size of a protein-binding site (3−10 nm) are examined using an electricfield-induced translocation of single molecules through a nanometer diameter pore. DNA, immersed in an electrolyte, is forced through synthetic pores ranging from 0.5 to 1.5 nm in radius in a 10 nm thick Si3N4 membrane using an electric field. To account for the stretching and bending, we use molecular dynamics to simulate the translocation. We have found a threshold for translocation that depends on both the dimensions of the pore and the applied transmembrane bias. The voltage threshold coincides with the stretching transition that occurs in double-stranded DNA near 60 pN.

The encyclopedic information encoded in the genome is packaged, copied, and transcribed through protein mechanisms that exploit the distinctive physical properties of DNA. Recently, single molecule spectroscopies have been developed to infer the average elastic properties of long (48 kbp ∼ 16.3 µm) molecules from force-extension measurements. Bustamante et al.1,2 have successfully fit these data over a range of forces (from 10 fN to 100 pN) using a simple wormlike chain model with the link length and extension modulus as parameters. When subjected to a force >10 pN, DNA deforms, stretching the double helix. It is remarkable that near 60 pN a dramatic transition occurssthe double helix stretches from 107% to nearly 170% of the length of B-DNA over a few piconewtons.3,4 The distance between bases in the stretched DNA is 0.7 nm, over 200% of the normal length. Despite the success of this model to capture the microscopic physics required to understand DNA-protein interactions, measurements of the mechanical properties of DNA over segments comparable to the size of a protein-binding site, which is about 10-30 base-pairs (bp) long (3-10 nm), are required. We propose to test the elastic properties of DNA on this length scale by using the field-induced translocation of DNA across a membrane through a nanopore.5-8 When an electric field is applied across the membrane with a pore in it, DNA immersed in electrolyte is attracted to the pore, blocks the current through it, and eventually permeates the membrane through the pore.5-8 The forces acting on single nucleotides in a DNA strand during the translocation are * Corresponding author. E-mail: [email protected]. † Beckman Institute. ‡ Department of Biochemistry. 10.1021/nl0510816 CCC: $30.25 Published on Web 09/08/2005

© 2005 American Chemical Society

approximately F ∼ qZ∆V/a, where qZ is the effective charge per nucleotide and ∆V is the electrostatic potential drop over the nucleotide separation a. Assuming that screening of the DNA charges by counterions is negligible, i.e., Z ≈ 1, and a ) 0.34 nm, the force on a single nucleotide corresponding to the applied voltage can range from 0 < F < 1 nN, depending on the distribution of the electrostatic field in a nanopore and the location of the nucleotide. As a result of this force, it has already been shown that folds and bends can occur in DNA during translocation through a 10 nm diameter pore, which affects the current and the translocation kinetics.9 In this paper, we report the results of a stringent test of the nanometer-scale mechanical properties of single DNA molecules. Single molecules of double-stranded DNA (dsDNA) and single-stranded DNA (ssDNA), immersed in electrolyte, were forced through synthetic nanopores, ranging from 0.5 to 1.5 nm in radius, in a 10 nm thick Si3N4 membrane using an electric field. Molecular dynamics (MD) simulations provided a comprehensive account of the bending and stretching that accompanies translocation of DNA. MD identified a threshold for permeation of dsDNA through the pore that depends on the radius of the pore and the applied transmembrane bias. These predictions were confirmed experimentally. Through MD simulations, we estimated the forces exerted by the electrostatic field on the DNA nucleotides at various depths in the nanopore and found that for dsDNA a stretching transition near a 60 pN tension determines the voltage threshold for permeation. We formed membranes using conventional semiconductor fabrication practices described elsewhere.10 Subsequently, a single nanopore is created in the membrane by electron beam

Figure 1. (Top) TEM images of pores with apparent radii of 0.5 ( 0.1 (a), 0.95 ( 0.1 (b), and 1.5 ( 0.1 nm (c) in a 10 nm thick Si3N4 membrane at 0° tilt. (Middle) Gel arrays showing five horizontal lanes with fluorescent bands that identify 58-mer ssDNA and 58 bp dsDNA at the positive (+) and negative (-) electrodes separated by 0.5 (e), 1.0 (f), and 1.5 nm (g) pores (with 100 bp ladder as a reference). The applied voltage was 200 mV. (Bottom, part d) I-V characteristics in 1 M KCl solution of the same three pores. Line fits yield slopes of 0.63 ( 0.03, 1.09 ( 0.03, and 1.24 ( 0.03 nS for the three pores, respectively.

stimulated decomposition and sputtering using a JEOL 2010F transmission electron microscope (TEM) operating at 200 keV. Figure 1a represents an array of TEM images taken at a tilt angle of 0° of pores with apparent radii of 0.5 ( 0.1 nm, 0.95 ( 0.1 nm, and 1.5 ( 0.1 nm in a Si3N4 membrane nominally 10 nm thick. (To account for elliptical shape, the radius Rp is taken to be the geometric mean of the major and minor axis of the pore.) The thickness of similarly processed membranes was determined to be 10 ( 2 nm using scanning electron microscopy and 11 ( 3 nm using electron energy loss spectroscopy. The TEM images represent a twodimensional projection through the membrane. The shot noise observed in the central area identified with the pore indicates perfect transmission of electrons through the membrane in that area. To investigate the three-dimensional structure, we tilted the membrane in the TEM about the pore axis and explored various defocus conditions. A model of the pore geometry that is consistent with these data is shown in Ho et al.10 and Heng et al.11stwo intersecting cones with a 10° cone anglesis represented schematically in the inset in Figure 1d. To characterize molecular transport, we measured the electrolytic current through a pore as a function of the applied electrochemical potential at 23.5 ( 1 °C. The currents were measured using an Axopatch 200B amplifier in resistive feedback mode with a 100 kHz bandwidth. All experiments 1884

were carried out in a membrane transport bicell made of either acrylic or PDMS, where the membrane separates two identical compartments each containing ≈40 µLs1 mL of 1 M KCl electrolyte and a Ag/AgCl electrode. Figure 1(bottom) shows the current-voltage (I-V) characteristics of the pores shown in Figure 1(top), measured over a range of (1 V in 1 M KCl after >55 h of immersion in deionized water. In all cases the I-V plots are approximately linear. A line fit to the data yields the conductance 0.63 ( 0.03 nS, 1.09 ( 0.03 nS, and 1.24 ( 0.03 nS. It is apparent that the conductance does not scale linearly with the radius derived from TEM. We attribute the observed variations in conductance to a fixed (negative) charge in the pore.10 Following the conductance measurements, DNA in 10 mM TRIS-Cl buffer at pH 8.5 was injected at the negative electrode. A voltage was applied across the membrane, and the current through the nanopore was measured. We have previously shown that under these conditions voltage-driven translocations of single molecules of DNA across the membrane temporary block the electrolytic current through the pore.7 Using the pore as a stochastic sensor, we have also shown that the duration of the current transients can be used to discriminate between ssDNA and dsDNA, as well as dsDNA of various length.7 Molecular dynamics simulations of DNA translocations through a synthetic nanopore12 predict a translocation velocity of 400 nucleotides/µs for an electric field of 1.4 V/5.2 nm, much faster than in the transmembrane pore of R-hemolysin, where the translocation velocity at a similar field is 10 nucleotides/µs.13 The narrow bandwidth (100 kHz) of the measurement apparatus used in these experiments limits the observation to transients longer than 10-100 µs. And even if the transient can be measured, it cannot be used to unambiguously discriminate between a translocation of DNA across the membrane and a temporary occlusion of the pore without a subsequent translocation.7,12 Therefore, to establish unequivocally that DNA permeates a pore and to count the number of copies, the minute amount of the DNA near the positive electrode was amplified using PCR and analyzed by gel electrophoresis or by quantitative PCR. This same methodology was used by Kasianowicz5 to first establish the correspondence between the number of current transients and the number of ssDNA molecules translocating through the pore of R-hemolysin. We tested the permeability of ssDNA (58-mer) and dsDNA (58 bp) through the pores shown in parts a-c of Figure 1 by injected 40 µg of DNA into the compartment housing the negative electrode and applying a voltage of 200 mV across the membrane for 4 h. The 58 bp dsDNA is obtained from the plasmid PUC19 (New England Biolabs, MA) cleaved at the EcoRI and HindIII sites, while the 58mer ssDNA (Midland, TX) has the same sequence as the 5′ strand of the 58 bp dsDNA. Typical gel arrays are shown in parts e-g of Figure 1 along with a molecular weight (MW) reference denoted as “100 bp ladder”, which contains a spread of DNA MW. After being stained, the DNA in each lane can be seen as bands spread according to the length (or MW). Notice that the ssDNA collected at the positive electrode, denoted by (+)58-mer, is observed to permeate Nano Lett., Vol. 5, No. 10, 2005

through all three pores and gives the same amplified pattern as the ssDNA injected at the negative electrode, denoted by (-)58-mer. On the other hand, dsDNA only permeates the Rp ) 1.5 ( 0.1 nm pore at 200 mV. We constructed a microscopic model of the experiment and carried out MD simulations of electrophoretic transport through pores. A crystalline Si3N4 membrane was built by replicating a unit cell of β-Si3N4 crystal along the unit cell vectors, producing a hexagonal patch of 10.3 nm thickness and 4.6 nm sides. By removing atoms from the Si3N4 patches, we produced pores of symmetric double-conical (hourglass) shapes with radii that correspond to our experiments. A DNA double helix with the same sequence used in our experiments was built from individual base pairs in the geometry suggested by Quanta.14 First, the DNA was placed in front of the pore, normal to the membrane, and then the Si3N4/ DNA complex was solvated in preequilibrated TIP3P water molecules. K+ and Cl- ions were added corresponding to a 1 M concentration. The final system measured about 30.3 nm in the direction normal to the membrane and included about 160 000 atoms. To simulate the system, the molecular force field describing water, ions, and nucleic acids15 was combined with the MSXX molecular force field for Si3N4.16 The protocols are described elsewhere.12 While the setup of our simulations faithfully reproduces the experimental conditions, nevertheless our simulations have the following limitations. First, the time scale of MD simulations is currently limited to a few tens of nanoseconds, which does not allow us to investigate processes that occur on a longer time scale. Second, our description of the silicon nitride surface is approximate, and more work has to be done in order to develop a microscopic force field that would quantitatively reproduce the binding energies between DNA and a silicon nitride surface. Because of the double helical structure of dsDNA, binding of DNA nucleotides to the surface of the pore is not likely and, therefore, the latter limitation does not prevent us from drawing quantitative conclusions about the conditions for dsDNA permeation. Finally, this study provides a stringent test of the AMBER95 force field that has been deployed in all our simulations. As shown below, we observed excellent agreement between simulation and experiment, suggesting that this force field accurately describes mechanical properties of dsDNA. At first, we simulated the 0.7, 1.0, and 1.5 nm radii pores combined with one 58-bp dsDNA helix. To shorten the translocation times to a practical duration, the simulations were done at 1.3 V. We applied a uniform electric field across a 10.3 nm thick Si3N4 membrane to drive a 58 bp dsDNA through 0.7 and 1.0 nm radius pores. The field induced a rearrangement of the ions and water that, in turn, focused the field to the vicinity of the membrane abolishing the gradient in the bulk and producing a 1.3 V transmembrane bias. As a reference point for DNA movement through a pore, we use the DNA center of mass (CoM) defined as ∑(zimi)/∑mi, where mi and zi are the mass and the zcoordinate of atom i, and the sums run over all DNA atoms. Although describing the conformation of a 58-bp DNA helix using a single variable can be ambiguous, in our case, CoM Nano Lett., Vol. 5, No. 10, 2005

faithfully illustrates the DNA permeation through a pore: when DNA fails to permeate the pore, its conformation is similar to that of a canonical double helix, whereas when DNA permeates the pore, the changes of the CoM position are much bigger than the shift of the CoM due to stretching, bending, or partial unzipping of the double helix. Figure 2a shows the position of the dsDNA CoM relative to the center of the Si3N4 membrane against time. Within the first few nanoseconds, the electric field drove dsDNA into the pores; the wider pore facilitated faster translocation. Due to their conical shapes, both pores narrow toward the center of the membrane, which slowed DNA translocation. After about 4.5 ns, the translocation halted in both simulations before the DNA arrived at the narrowest part of the pores. The snapshots included in the inset illustrate the conformations of dsDNA at the end of these simulations. Notice that dsDNA traveled a longer path inside the wider pore suggesting that the translocation halts at a particular cross section. To ascertain if the translocation halted at the same cross section in both pores, we plot in the inset of Figure 2a the local radius of the pore around the first DNA base pair. In both simulations, the translocation of dsDNA halted when the pore narrowed to 1.25 ( 0.1 nm radius. So, we conclude that pores with Rp < 1.25 nm are impermeable to dsDNA for fields 2.6 V. The snapshot illustrates the conformation of dsDNA at 3.2 V after 12 ns. Animations illustrating these MD simulations are available at http:// www.ks.uiuc.edu/Research/nanopore/stretching/.

the pore, deforming the double helix of DNA into a loop, as shown in Figure 2b. A translocation like this was reported earlier.12 1886

To establish the threshold field and the force required to drive dsDNA through the pore, we restarted our simulation of the 1.0 nm radius pore from the conformation shown in the inset of Figure 2a. In Figure 2c we plot the dsDNA center-of-mass against time for biases ranging from 1.3 to 6.5 V. These results indicate that dsDNA permeates a 1.0 nm radius pore for voltages >2.6 V. In all MD simulations, we observed partial unzipping of the DNA ends when they pass by the narrowest part of the pore. The length of the unzipped fragment was found to vary from one simulation to the other, but typically was from 2 to 12 nucleotides. The 3.9 V path in Figure 2c is longer than the 3.2 V path because at 3.9 V, dsDNA partially unzipped before reaching the center of the membrane. After both DNA strands had passed through the center of the membrane (12.5 ns), the translocation at 3.9 V proceeded faster than that at 3.2 V, as indicated by the slopes of the curves. As our simulations are essentially in silico single molecule experiments, such variation of the outcome is expected. Due to the limited resources, we could carry out only two additional simulations to check the reproducibility of our results. In the first simulation, carried out at 2.6 V, dsDNA failed to permeate the pore, whereas in the second simulations (at 3.2 V), we observed translocation of dsDNA through the pore, confirming our estimated of the translocation threshold. The predicted thresholds were tested experimentally using two pores with 1.0 ( 0.1 and 1.1 ( 0.1 nm radii. Forty micrograms of 58 bp dsDNA was injected into the compartment housing the negative electrode, and a trans-membrane voltage is applied across the membrane for 4 h. The gel array in Figure 3a indicates that 58 bp dsDNA can be forced through a 1 nm radius pore if the applied voltage is >2.75 V. Notice that we only observe a fluorescent band in the lane corresponding to the positive electrode, (+)58 bp, for an applied bias of 3 V. We obtained a similar result for a 1.1 nm radius pore, but the observed threshold voltage was lower: 58 bp dsDNA permeates the pore for V > 2.25 V. We also measured the number of DNA copies translocating near threshold by quantitative real-time PCR. To conduct the qPCR experiment, we use 622 bp dsDNA obtained from the plasmid pNF-kB-Luc (Stratagene, CA) cleaved at the two EcoRI sites. Two PCR primers were designed to amplify a 72 bp region within a 622 base target sequence. A TaqMan probe was designed, mapping to this 72 bp target sequence and labeled with an FAM reporter dye at the 5′-end and with a TAMRA quencher dye at the 3′-end. During amplification, the primers and the probe hybridize to the target sequence. Subsequently, during polymerization, the 5′ exonuclease activity of the DNA polymerase cleaves the probe releasing the reporter dye and resulting in increased fluorescence, which is proportional to the amount of target. The same experiment is repeated by injecting 40 µg of 622 bp dsDNA into the negative electrode compartment electrode and applying voltage bias across the membrane for 4 h. Parts b and c of Figure 3 represent the results of qPCR analyses showing the number of copies found to permeate through the 1.0 and 1.1 nm radius pores as a function of the applied potential. In correspondence with the 2.75 V threshold Nano Lett., Vol. 5, No. 10, 2005

Figure 3. (a) Gel arrays containing three horizontal lanes with fluorescent bands indicating 58 bp dsDNA found at the negative (-) and positive (+) electrodes in a bicell with a 1 nm radius pore (along with a 100 bp ladder.) The 58 bp permeates the pore only for V > 2.75 V. (b) Similar to (a) but instead using a 1.1 nm radius pore. The 58 bp permeates this pore only for V > 2.25 V. (c) qPCR results obtained for the same pores showing the copy number versus voltage. (upper left insets) TEM micrograph taken at 0° tilt for 1 nm radius (blue) and 1.1 nm radius (red). 622 bp dsDNA permeates the 1 nm pore for V > 2.5 V, and the 1.1 nm pore for V > 2.0 V. (lower right inset) Expanded view of the data near threshold.

obtained from the gel array for 58 bp, the 1.0 nm pore exhibits a threshold voltage of 2.5 V for 622 bp dsDNA. Similarly, the 1.1 nm radius pore shows a threshold voltage of 2 V. The 0.25 V discrepancy observed between the gel array and qPCR data may be attributed to the difficulties associated with capturing and handling such a small number of DNA copies. To elucidate the origin of the sharp voltage threshold for permeation of dsDNA, we examined the distribution of the electrostatic potential inside a 1.0 nm radius pore. Applying a 1.3, 2.6, 3.2, 3.9 and 6.5 V bias, we simulated the nanopore system without DNA. For each value of the transmembrane bias, the distribution of the electrostatic field was determined by averaging instantaneous snapshots of the electrostatic potential over the corresponding MD trajectory; the method for computing such average electrostatic potential maps has been described elsewhere.18 The distribution of the electrostatic potential in a 1.0 nm radius pore at a 2.6 V bias is shown in Figure 4a. The potential drops abruptly near the pore’s constriction. A similar distribution is observed when dsDNA blocks the pore, Figure 4b, although the influence of the DNA’s charge on the distribution is apparent. A representative conformation of dsDNA in this pore at a 2.6 V bias is shown in Figure 4c. Figure 4d shows the force Nano Lett., Vol. 5, No. 10, 2005

Figure 4. Distribution of the electrostatic force and potential in a nanopore. (a) and (b) Electrostatic potential contours inside a 1.0 nm radius pore at 2.6 V without (a) and with (b) DNA present in the pore. (c) A snapshot from the simulation corresponding to the electrostatic potential contour shown in panel b. (d) The force versus distance from the center of the pore in a 1.0 nm radius pore that does not contain DNA. (Inset) Expanded view of the same data near the pore constriction. The vertical bands indicate the location of the leading DNA base pair in the simulations illustrated by Figure 3c. (e) Computed directly from MD trajectories, the total nonbonded force acting on DNA base pairs in a nanopore at a 1.3, 2.6, and 3.2 V bias.

acting on a probe charge (-1.6 × 10-19 C) in an empty pore (no DNA), determined by differentiating along the axis of symmetry of the pore the distribution of the electrostatic potential. At every bias, the force increases rapidly near the pore center, the maximum force scales with the transmembrane potential. The inset to Figure 4d shows the distribution of the force near the DNA leading edge which position is 1887

indicated by the vertical rectangle at 2.6 (no permeation) and 3.2 (permeation) V. At 2.6 V, we estimate the average force acting on a nucleotide at the leading edge of DNA at about 45 pN, while above the threshold (3.2 V) this force exceeds 60 pN, sufficient to stretch the helix and allow the molecule to pass. The above estimates of the force are based on the distribution of the electrostatic potential in the pore without DNA in it. The presence of DNA and the counterions screening the DNA charge could conceivably reduce this force. Visual examination of MD trajectories revealed that, although a cloud of counterions surrounds DNA, the ions do not stay bound to DNA, exchanging their positions with the neighbors and producing the ionic current. To determine the force acting on DNA, we computed a nonbonded force (van der Waals and electrostatic) on every DNA atom, and averaged this force over the MD trajectory and over the atoms comprising the base pairs. The resulting force is plotted in Figure 4e for a 1.3, 2.6, and 3.2 V bias. The values of the force are similar to the estimates obtained using the distribution of the electrostatic potential in an empty pore: at a 3.2 V bias, the force exceeds 60 pN near the pore constriction, whereas it vanishes away from the membrane. Thus, the screening of the electrostatic force by the counterions in our pores is small when the transmembrane bias exceeds 1 V. In conclusion, we have demonstrated that, using an electrostatic field in a nanopore, one can exert forces on DNA in the range of 1-300 pN and observe the threshold for dsDNA permeation conditioned by mechanical properties of dsDNA. It might become possible in the future to use the electrostatic force in a nanopore to probe mechanical properties of DNA-protein assemblies. Acknowledgment. We gratefully acknowledge frequent discussions with Jean-Pierre Leburton and Narayana Aluru in the Beckman Institute, and the use of the Center for

1888

Microanalysis of Materials supported by a U.S. Department of Energy grant (DEFG02-91-ER45439). We also thank Dr. Susan Power of Cellular Molecular Technologies Inc for her help in conducting the qPCR experiments. This work was funded by grants from the National Science Foundation (0210843), the Air Force (FA9550-04-1-0214), and the National Institutes of Health (P41-RR05969 and R01-HG00371301); computer time was provided through a National Resources Allocation Committee grant (MCA93S028). References (1) Bustamante, C.; Smith, S.; Liphardt, J.; Smith, D. Curr. Opin. Struct. Biol. 2000, 10, 279. (2) Bouchiat, C.; Wang, M. D.; Allemand, J.-F.; Strick, T.; Block, S. M.; Croquette, V. Biophys. J. 1999, 76, 409. (3) Cluzel-Schaumann, H.; Rief, M.; Tolksdorf, C.; Gaub, H. E. Biophys. J. 2000, 78, 1997. (4) Smith, S. B.; Y. Cui, Bustamante, C. Science 1996, 271, 795. (5) Kasianowicz, J. J.; Brandin, E.; Branton, D.; Deamer, D. W. Proc. Natl. Acad. Sci. U.S.A. 1996, 93, 13770. (6) Meller, A.; et al. Phys. ReV. Lett. 2001, 86, 3435. (7) Heng, J. B.; Ho, C.; Kim, T.; Timp, R.; Aksimentiev, A.; Grinkova, Y. V.; Sligar, S.; Schulten, K.; Timp, G. Biophys. J. 2004, 87, 2905. (8) Akeson, M.; Branton, D.; Kasianowicz, J. J.; Brandin, E.; Deamer, D. W. Biophys. J. 1999, 77, 3227. (9) Li, J.; Gersho, M.; Stein, D.; Brandin, E.; Aziz, M. J.; Golovchenko, J. A. Nat. Mater. 2003, 2, 611. (10) Ho, C.; Qiao, R.; Heng, J. B.; Chatterjee, A.; Timp, R.; Aluru, N.; Timp, G. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 10445-10450. (11) Heng, J.; Aksimentiev, A.; Ho, C.; Dimitrov, V.; Sorsch, T. W.; Miner, J. F.; Mansfield, W. M.; Schulten, K.; Timp, G. Bell Syst. Tech. J., in press. (12) Aksimentiev, A.; et al. Biophys. J. 2004, 87, 2086. (13) Mathe, J.; Aksimentiev, A.; Nelson, D. R.; Schulten, K.; Meller, A. Proc. Natl. Acad. Sci. U.S.A. 2005, 102, 12377-12382. (14) Polygen, Quanta Polygen Corporation, 1988. (15) Cornell, W. D.; Cieplak, P.; Bayly, C. I.; Gould, I. R.; Merz, K. M., Jr.; Ferguson, D. M.; Spellmeyer, D. C.; Fox, T.; Caldwell, J. W.; Kollman, P. A. J. Am. Chem. Soc. 1995, 117, 5179. (16) Wendel, J. A.; Goddard, W. J. Chem. Phys. 1992, 97, 5048. (17) Yeh, I.; Hummer, G. Proc. Natl. Acad. Sci. U.S.A. 2004, 101, 12177. (18) Aksimentiev, A.; Schulten, K. Biophys. J. 2005, 88, 3745.

NL0510816

Nano Lett., Vol. 5, No. 10, 2005